As someone who has the official dealer diagnostic tool and has spent some time analysing the workings of the battery management system in these cars, has 3 fairly rapidly failing cells and some second hand replacement cells to fit quite soon, this thread is very interesting.
Your use of 18650 cells makes me a little nervous to be honest, and one question I didn't see addressed in your description is how did you deal with getting a thermal bond from the cells to the temperature sensor on the board ? Without correct temperature information from the monitored cells (only 3 out of every 4 cells is monitored) the rapid charge rate may be severely limited as charge rate will be limited by both low and high temperatures, and with no sensor on the bodged cells they could potentially overheat without the BMS being aware, which in extreme cases could lead to a fire.
Even with an attached temperature sensor, the BMS is programmed with knowledge of the thermal stability of the original cells used - which are a chemistry which is quite tolerant of heat, at least from a not catching fire perspective. More modern higher density cells are often less thermally stable and not suitable for high output without thermal management so the BMS might allow them to be pushed outside their safe operating range which again could result in a fire.
Not trying to poke holes, just warning you of some of the potential risks of your bodged 18650 cells, or the other prismatic cell you used for that matter, especially in Australian heat, where these cars are not well suited in the first place due to lack of active temperature management of the cells when Level 2 charging or driving. (Only during DC rapid charging is the battery pack cooled)
ThisNonsense wrote:Issues encountered along the way
I spoke a little earlier about having an issue with the balancing. Basically, one of the new cells is not being correctly balanced. It's not out by much, but the cell is consistently 0.035v higher than the other replaced cell, and is always the first cell to reach 4.10v (full charge as far as the BMU is concerned). This is mainly an issue because the two new 40Ah cells aren't really giving me that capacity (more on this later too). I think maybe the balance lead has come loose (it wouldn't surprise me if I hadn't tightened it properly when we first assembled the pack) however I haven't had a chance to drop the pack again since I've started getting suspicious about how this cell is out of balance.
I think you'll find the "balance lead" and the voltage sensing lead are one and the same. So the balance lead can't become detached without the LTC chip then being unable to read the cell voltage, and you'd soon know about that as it would put an error on the dashboard and refuse to let you charge...
What's more likely happening is that you just have too much capacity difference between the best and worst cells. These cars do top balancing but we think they also do a little bit of bottom balancing sometimes as well. If you start charging below about 3 bars you will sometimes get a "pause" where the charge power drops down to near nothing for 15-30 minutes.
Although I haven't seen conclusive evidence, we think that there is bleed only balancing occurring at that point. The problem with this is that if there is a large variation in capacity between cells any sort of bottom balancing (attempting to equalise cell voltages at a low SoC) will cause the balance state at a high SoC to get even further out. If the cell capacity variation is too great then it will be impossible to reach 100% SoC with balanced cell voltages.
One way you could test this theory is to put the car through several discharge/recharge cycles while only going down to about 50% SoC - this is high enough that bottom balancing will not occur, and if the reason it is having difficulty balancing is due to a very low capacity cell after a few half cycles like this it should reach balance at full charge. (However it will start going out again if you discharge it low)
The first thing I'll mention is that so far everything has been going well (since I replaced the bolts and the balance boards we killed). The cells have held up fine and there doesn't appear to be any degradation going on with the new cells from their treatment. Unfortunately, they're also not giving me anything near 40Ah. In fact I seem to be getting something close to 20Ah, but the BMU is taking a while to adjust down. I think there are a couple of reasons this might be happening.
Firstly, I'm a little suspicious of the rating given on the Taobao listing. I wouldn't be surprised if I've bought cells that are 30Ah at 1C discharge and 40Ah at 1 amp or something like that.
Secondly, since the iMiev BMUs only work from 3.60v to 4.10v and the capacities listed for the battery seems to be measured from 2.75v to 4.20v I'm pretty sure there's going to be about 10% to 20% loss in "usable capacity" as measured by the car.
Correct - you'll lose around 20% of the usable capacity by reducing the operating range from 4.2-2.75 to 4.1 to approx 3.6 which is used in the car. So your replacement cell would only be 40Ah over that full voltage range so perhaps 32Ah at most in use in the car, even if it lived up to its claims.
I went through a similar thought process when I obtained four second hand cells and wanted to test their usable capacity - what mattered to me is the usable Ah capacity over the voltage range the car use them, so that I could compare their performance to the other cells in the car to ensure the ones I was fitting were actually better than the ones I was replacing...
So I did a discharge test from 4.1 volts down to 3.63 volts (which seems to be 0% SoC on my car - it varies slightly with the SoH of the battery) and then also down to 3.0 volts, which is the lowest my discharger will let me take the cells.
And on all four I got 40-41Ah for the 4.1v to 3.63v test, and about 43-44Ah for the 4.1v to 3.0v test. With a full 4.2v to 2.75v test I'd say they'd probably have got 48Ah.
I think the reason why the car will only take the cells down to about 3.6 volts open circuit is simple - at that SoC the cells can still put out the full 150 amps needed for full acceleration without dipping below 3.0 volts. If you were to discharge them down to say 3.2 volts then acceleration would have to be severely limited to prevent the cells going below 3.0 volts under load, so I think the chosen cutoff voltage is actually based on >3.0 volts under a 150 amp load.
If any cell dips below 3.0 volts at any time the turtle light will come on and acceleration will be severely restricted. (Ask me how I know...
Thirdly, I'm being held back a little by one of the cells hitting full charge first and this meaning the other cell reaches empty first. Once I get a chance to check the balancing lead I'll have to see if that makes any sort of noticeable difference.
Are you seeing this on 3kW Level 2 charging as well, or only on DC rapid charging ? If you're referring to rapid charging this is a sure sign that you have a cell with high internal resistance, especially if the cell goes from being lower voltage than the others before the charge to higher than the others very soon after starting charging.
My three faulty cells all have high internal resistance, one is particularly bad - as a result during rapid charging they hit 4.105 volts a long time before the others and cause premature throttling of the rapid charging speeds. Even if I plug in in summer as low as 20% SoC the weak cells will reach 4.105 volts within about a minute and start tapering the charge rate despite the car still only being at about 25% SoC! It used to charge at a full 43kW up to nearly 50% SoC. I find this just as annoying as the range loss I've had.
Rapid charging times from 20% to 80% have gone up from around 17 minutes to nearly 30 minutes just due to this high cell resistance. In summer. Winter is even slower as cell internal resistance goes up in cold cells.
Finally, I think there's a calibration issue either caused by trying to voltage match the cells before they were installed (so the BMU isn't sure how much capacity it actually has any more) or by a difference in voltage curves for different lithium battery chemistries. The calibration issue makes some sense to me because when I took the old cells out, they were very nearly dead and nearly a third of a volt lower than the rest of the cells. When I installed the new cells I had to put about 10-15Ah into them to bring them as close to being in balance with the rest of the cells as I could. This might mean that the BMU doesn't know about those extra Ah at the bottom.
If you replace faulty cells in the pack and thus increase their usable capacity then unfortunately the BMS will not immediately be aware of the change and will refuse to let you use the additional capacity that is now there. It has a record of the usable Ah capacity of the pack (the figure reported by Canion/Hobdrive) and will start counting that Ah capacity down from 100% as you discharge the battery and when it reaches what it believes is about 10% SoC you will go into turtle mode regardless of how much capacity is really left. And when it reaches what it believes is 0% the car will shut off, even if the cell voltages are still quite high.
If you drive the car many hundreds of miles over many many full charge discharge cycles it will eventually start to make incremental upwards revisions to the usable capacity figure - see jray3's battery swap thread
for an example of this. It will only increase by approx 1Ah per thousand miles, so if you're replacing a really sick cell with a good one it could take a very long time and you may never get the full usable capacity.
The dealer diagnostic tool has two routines that can be run when cell/pack swaps are performed - one is just a factory reset of the ECU which goes back to an assumed new 45.8Ah usable capacity - this is used if the dealer fits a brand new (less than 6 month old) replacement pack, so not usable for old packs.
The other routine is a battery calibration. This basically tells you to discharge the battery down to about 3.75 volts per cell then fully charges the battery while under supervision of the diagnostic tool. At the end of that a new Ah figure based on an actual measurement is made and recorded in the BMS memory, and it also records a figure for the best and worst cells in the pack. This method "instantly" updates the Ah figure to the actual true state of the battery.
I have tried this battery calibration routine a couple of times on my car in the last year and unfortunately both times it caused the reported capacity to go DOWN because the faulty cells were degrading faster than the BMS estimated so it's most recent figure was always behind the true capacity of the cells... after I replace the faulty cells I will do a calibration again and the figure should go UP this time...
You really want to try to get your hands on a diagnostic tool that can do the battery calibration for you if you want to have the full usable capacity of the repaired battery available. I would have thought there were plenty of Mitsubishi's in Australia - any authorised dealer should have a MUT3 diagnostic tool as it does all models of Mitsubishi.
I've noticed my car seem to go into turtle mode when the last cell hits about 3.67v under no load (maybe around ~3.5v under full load) and thinks it has about 2.5Ah remaining, both of those seem a little wrong to me.
No, that sounds about right to me, or at least very close to how my car behaves - both before I had cell degradation and after. If anything mine hits turtle at a slightly higher voltage. Last time I did a test the lowest cell being 3.63 volts corresponded to 0% SoC, and turtle mode kicks in with a SoC of around 10-12% so slightly higher voltage than this, probably around 3.7 volts.
Turtle mode is triggered by two different things - estimated SoC being below about 10-12%, or instantaneous cell voltages going below 3.0 volts. Normally the SoC trigger always happens first, unless you have a faulty cell or a wildly inaccurate Ah estimate in the BMS.
The other option is the difference in voltage curves, this one is based on the idea that the "Amp hours remaining" value given out by the BMU is calculated based on the voltage of the pack each time the car starts up and then is tracked based on current draw during driving until the car is started again, and then the process starts again. I might sound a bit odd, but I've noticed that if I charge the car to full from turtle mode (~2Ah remaining) I'll put in about 6kWh. This works out to be ~17Ah, but the readings from the BMU will say ~25Ah total capacity and ~22Ah remaining (meaning the car thinks I've put in an extra 3Ah). If I drive home and end up with ~15Ah remaining and everything seems fine, then I'll come back to my van the next day and suddenly find I'm down to ~9Ah remaining. There doesn't appear to be any self discharge occurring, all the cell voltages are still the same and it doesn't appear to drop in capacity if I drive around for a bit and then come back a day later. This change also shows up as a mysterious drop in range despite the car sitting stationary. (There's also a possibility that the program I'm using, HOBDrive, calculates the remaining capacity itself but that wouldn't explain the drop in range.)
I see these strange jumps in reported SoC a bit on my car too - but only since some cells have developed faults, and not as bad as you see it. For example I'll drive to work and arrive with 8 bars remaining, when I get into the car to drive home 8 hours later I actually have 9 bars remaining!
Or another one that happens a lot in winter is I'll drive down to about 40% SoC as reported by Canion, stop to rapid charge, the instant that I start rapid charging the reported SoC will drop to say 20%, and I'll lose two bars on the dashboard, and it will take 5 minutes of charging just to (supposedly) get back to where I was.
Then the rapid charge will end at say 75% reported instead of the usual 82%, but when I unplug the car and turn it on suddenly the SoC is reported as 82%.
I think the cause of this is large discrepancies in the capacity of cells between best and worst cells confusing the voltage estimation the BMS uses. As you say, while you are driving the car coloumb counting is used to keep track of SoC based on Ah in/out, however when you turn the car on (or start rapid charging it seems) and the car is near full or near empty and the voltage estimated SoC is significantly different than the columb counted figure it makes an "adjustment", however I think this adjustment process is confused when there is too much voltage spread between individual cells.
I certainly did not notice this curious behaviour back when all my cells were good, and I'm hoping it goes away after the cell swap.
So I would say this symptom is also caused by too much variation in cell capacity between your best and worst cells.
Other than that, I appear to be getting quite a few cell errors pop up when driving but they're very random as to when they occur. Sometimes I'll get one every minute or so, other times I'll get one or two every 30 minutes. The appear to be messages about individual cells being out of balance, but I can't remember their individual codes right now. Obviously I also get a pack balance error due to the new cells ending up being more than 0.10v out of balance with the rest eventually.
I can't remember it off hand but there is one particular error code that is set if the cell voltage variation between cells is greater than a certain amount and SoC is >50%. You're probably hitting this condition due to your cell balancing issues and again it will be down to too much variation in characteristics between your best and worst cells.
By the way, I don't know whether these guys would be willing to ship all the way to Australia from the UK or whether it would be prohibitively expensive, but if you're wanting to keep this car on the road for a while yet it might be worth investing in some genuine replacement cells:https://www.secondlife-evbatteries.com/ ... eries.html
I bought four cells from them and as mentioned earlier while second hand they are tested before sale and at least for the four I bought, their capacity was good, exceeding the performance of even the best cells in my pack let alone the bad ones I'm going to swap out.
It would sure solve a lot of problems compared to trying to bodge unknown quality mismatched cells into the car. It's always going to be a tough challenge to fit one odd cell that doesn't match the others in characteristics without upsetting the BMS and causing the array of side effects that you're seeing...